Splitting a check is normally framed as a fairness problem: who had the espresso martini, who only had water, how do you divide the tax. There is a quieter version of the question that has nothing to do with fairness, and it costs real money. When the check comes apart, the discount on it may not survive the trip.
This is not a rumor or a server error. It is written into the vendor documentation. Toast’s own support article for splitting checks by item says item-level discounts are removed from the item when it moves to a new check and have to be reapplied by hand. Square documents something different again — and the difference is the whole story, because it decides whether your discount moves with your food, sits on somebody else’s check, or simply stops applying.
The short answer to the title question: sometimes, and which time depends on two things you can actually check — what kind of discount it is, and which split the server performs. The rest of this is how to tell them apart before the check closes.
What actually happens when a check is split by item?
Mechanically, splitting by item is a move. The server opens a split screen, taps items, and drops them onto a new check. The items travel. The question is what travels with them — and the answer is that a line item is not a self-contained object. It carries a price, but the discount attached to it may be a separate record that points at the original check.
Toast is unusually direct about this in the support article its own restaurants read:
“Item-level discounts also do not carry over when you split a check — the discount is removed from the item once it moves to the new check, so you’ll need to reapply any discounts on each split check individually.”
Read that as a workflow instruction and it is unremarkable: staff are told to reapply discounts. Read it as a diner and it is a warning. The documented fact is narrow but sufficient — reapplication is manual, so the discount survives only if someone performs an extra step. How often that step gets missed is not something Toast publishes, and we are not going to pretend otherwise. The same document notes that service charges do not translate across split checks either — a related failure we cover in the piece on what a split sees at the swipe versus at the receipt, which digs into the terminal mechanics and the nine-way cap in depth.
The asymmetry that matters: if the discount is not reapplied, the amount owed goes up, and what you are looking for is a line that is no longer there. A wrong number invites a second look; a missing one does not. That is a point about reading a receipt, not a behaviour Toast documents.
Source: Split Checks on the POS, Toast
Square does it differently — and creates a different problem
It would be convenient if every point-of-sale behaved like Toast. They do not, and assuming they do is how you end up arguing with a server who is correctly telling you their system works another way.
Square’s support article on splitting a check states that when a check is split by seat, discounts, coupons and service charges that are applied to specific items, or configured to be item specific, move alongside the item they are associated with. So far, better news. But the same paragraph adds the catch:
“If there are any cart-level service charges, discounts, or promotions, then those would remain on the original check and can be moved manually.”
That is a different failure with a different shape. Nothing is documented as deleted here — the discount is stranded. A whole-table promotion stays attached to whichever check is the original, and the other checks are created without it. Square’s documentation stops there, so this page will too: it says where the discount stays, not what it is then worth or what the group’s combined total comes to. The part that matters at the table is simply that everyone on a newly created check is paying an undiscounted share unless someone moves the discount by hand.
Two different failures, one lesson. On Toast, an item-level discount is removed and has to be reapplied, so a group that does not notice pays more in total. On Square, a cart-level discount is not removed but is created only on the original check. Square does not say what the group’s combined total then comes to, so the safe reading is the narrow one: what each person owes stops matching what they expected. Different problems, same fix: know which one your table is exposed to before the check is divided.
Source: Split a check by item or seat, Square
The third case: the deal stops being true
There is a category of discount that does not get moved or stranded, but invalidated — because splitting the check destroys the condition the discount was testing.
Buy-one-get-one is the clean example. A BOGO is a rule about a pair of items on one check. Split those two items onto two checks and each check now holds a single item, so the rule no longer sees a qualifying pair. In a 2025 thread on Square’s seller community, a restaurant operator described exactly this: a weekly BOGO that “works quite naturally on the POS — until you split the items across seats,” after which the automatic discount stops applying.
The reply that the community marked as the solution — written by a Square Champion who states plainly that he is a fellow seller and not a Square employee, so treat it as a well-informed merchant account rather than vendor documentation — explains the mechanism: an even split only prorates the total and moves nothing, while splitting items across seats moves them into their own transactions, so “the automatic discount rules don’t recognize that the BOGO was satisfied.” A second operator replied months later insisting it was a regression rather than intended behaviour. Either way, the diner-facing result is the same.
What is actually established here is one vendor, one discount type, one merchant’s report. The reason it still matters is the mechanism rather than the sample: any conditional promotion — spend $50, get $10 off, buy two entrées and get an appetizer free, a prix-fixe requiring three courses — is a claim about a whole check, so a rule engine evaluating a smaller check could reach a different answer. Treat that as a question to ask rather than a documented outcome: nobody has published what each system does with each offer type, and we are not going to assert it for them.
Source: Automatic Discounts and Split Items Workaround, Square Seller Community, 2025 (merchant discussion, not official documentation)
Which discounts survive a split?
Ask one question before the check is divided: what is this discount attached to? The answer predicts the failure.
One caution about reading that table: each cell reports what a named vendor documents for a named operation, not a universal rule. Square’s note describes splitting by seat; Toast’s describes splitting by item. Other systems, and other split methods on these same systems, may behave differently and are simply not covered by either document.
That last row is the one people get backwards, and it is the exception that rewards attention: Toast’s documentation says that if guests want to earn loyalty points across multiple loyalty accounts on a single order, the check should be split rather than the payment — splitting the payment instead means points will not accrue properly. So for loyalty, asking for a genuine check split is the move that protects you. For item-level discounts on the same system, it is the move that endangers them.
Sources: Split Checks on the POS, Toast; Split a check by item or seat, Square
Is the discounted line really where pricing goes wrong?
Partly — and the research is more specific than the obvious version of this argument, in a way worth reporting honestly because some of it cuts the other way. The literature on checkout price accuracy does not find that discounts in general are error-prone. It finds that advertised specials are, while other kinds of price reduction are not.
The reference point is the FTC’s 1998 Price Check II — described by NIST as a larger, more comprehensive review than the 1996 survey that preceded it — run using the price-verification procedure NIST and the National Conference on Weights and Measures developed jointly, which 42 states had adopted. It compared scanned prices with the lowest posted or advertised price across 107,096 items. Overall accuracy was decent — one in 30 items was mispriced, and errors split roughly evenly between overcharges and undercharges. Sale items were the exception on both counts.
The FTC’s own consumer-protection director drew the practical conclusion at the time: consumers can protect themselves by paying close attention to the prices they are charged, particularly when buying sale items, because almost two-thirds of the scanner errors in that category were overcharges. The inspectors also catalogued the causes, and they are mundane and familiar — out-of-date signs, incorrect prices in the computer, shelf prices that never got updated. Several of those are failures to propagate a price change, which is a plausible reason advertised prices would be the fragile ones — though the inspectors catalogued causes rather than testing that explanation, so read it as a hypothesis, not a finding.
Sixteen years later, a team of pricing researchers went back to that same FTC dataset and re-ran it in the Journal of Retailing, this time measuring the size of each error rather than only its direction. The asymmetry held: undercharges ran about 22% for both sale and non-sale items, while average overcharges were 19% on non-sale items and 38% on sale items. The same 2014 paper sets those figures against the work that opened this line of research: it reports that Goodstein’s 1994 study in the Journal of Marketing found average overcharges of 6.9% against average undercharges of 23.7%. Worth flagging, in the same spirit as the caveats elsewhere on this page: Goodstein’s original 1994 article sits behind a publisher paywall we could not open directly, so those two figures reach this page as quoted inside the 2014 re-analysis, not independently re-verified against the 1994 text itself. On that basis, the 2014 authors’ reading is that overcharges have become more severe even as the overall rate of mistakes has fallen.
Average size of an overcharge on sale items versus non-sale items, in the FTC’s 1998 Price Check II data as re-analysed in the Journal of Retailing, measured as a share of the price charged. This is how big an error is once one has occurred — not how likely one is.
Now the part that argues against the easy version of this page’s claim, and which the same 2014 paper is explicit about. When its authors modelled promotional activity, only advertised sale items came out worse: they were associated with significantly greater error overall. Items on a temporary price reduction went the other way — errors and overcharges both declined significantly, which the authors state as a formal generalisation that errors are significantly less likely on temporarily reduced items. Permanently discounted items had no significant effect on error rates, and neither did items tied to store coupons.
So the honest version is narrower: “discounts are fragile” is not what this research shows. Where errors did cluster was on items whose price had been advertised — an association the authors report, not a mechanism they test. In that study the advertising was a supermarket flyer, which is not a happy-hour board, and the researchers made no claim about restaurants. The resemblance is the reason the parallel is worth drawing at all — and the difference is the reason it stays a parallel rather than evidence.
The overcharge asymmetry on sale items does replicate outside the FTC data. The same 2014 paper obtained a separate set from Washington State — the last quarter of 2005, 100 items checked in each of 559 stores — and found the same direction there: controlling for the price of the product, relative errors were greater for sale overcharges (36.39%) than for sale undercharges (26.46%). Worth knowing how that is measured, because the wording invites a different reading: these are error sizes expressed as a percentage of the price charged, not the share of items priced wrongly.
Second, and more important: this literature examines retail and grocery checkout scanning, not restaurant point-of-sale check splitting. Nobody has run a Price Check study on split restaurant checks, and none of these findings measure what a POS does to a discount when a server moves an item. They are context for why an advertised price is a plausible weak point, not evidence about splitting. The claim that a split check strips a discount rests entirely on the Toast and Square documentation above — which is a much stronger kind of evidence for that specific claim anyway, because it is the vendors describing their own systems.
Sources: Price Check II, NIST/FTC, 1998; Price Verification FAQs, NIST; The Accuracy of Scanned Prices, Journal of Retailing 90(2), 2014; UPC Scanner Pricing Systems: Are They Accurate?, Journal of Marketing 58(2), 1994
What is actually at stake when a discount goes missing?
Percentages are abstract until they sit on a real total. Two thirds — 66% — of splitty’s US-leaning scanned restaurant receipts come to $100 or more, and nearly half — 48% — come to $150 or more (August 2026 snapshot of splitty’s own scanned receipts; percentages only, per our first-party data policy). Those are checks large enough that a percentage-off promotion is worth real money — though how much depends on the size of the discount, which these receipts do not record.
The illustration below uses invented round numbers to show the shape of the loss, not measured ones. Four people at happy hour, four discounted drinks at $6 off each, and a server who splits the check by item on a system that drops item-level discounts:
Notice what each person would have to do to catch it. Their own check is internally correct: their drink is on it, at the menu price. Spotting the loss requires remembering what the discounted price was supposed to be, or comparing against a combined check that has now been replaced by four separate ones. That is a claim about what the documents make easy to check, not a prediction about how observant any particular table is.
How to protect a discount through a split
Do these before the check closes. A closed check is a harder conversation than an open one, and on most systems it is a different operation for the staff.
Before any split, get the check with everything on it and take a picture. That image is your record of what the discount was worth, and it is what you will point at if the split checks do not match it. This is the same habit that protects you when the bill has errors on it.
On Toast, reapplying an item-level discount after a split is a manual step, so it happens only if someone does it. Naming it turns it into a step that was asked for: “We’ve got the happy-hour price on four of these — can you make sure that carries onto each check?”
Each check will be internally consistent, so look for the discount line itself rather than trusting a total. Totals are a weak test in both directions: they can move for reasons that have nothing to do with a discount, such as tax recalculation or rounding, and a discount stranded on the original check can leave the group’s combined total unchanged while individual shares are wrong.
BOGO, spend-and-save and bundle offers are rules about a whole check, and at least one Square merchant has reported a BOGO dropping once the qualifying items were separated. Whether yours survives is a question for the restaurant. If the answer is no, keeping the check whole and settling up between yourselves is the safe route.
Square documents that a cart-level promotion stays on the original check and can be moved manually. So ask which check kept it, and either have staff redistribute it or account for it yourselves when you settle up.
What to say at the table
Short, specific and unembarrassed works better than apologetic. None of these accuse anyone of anything — they are requests about a documented system behaviour.
“Before you split it — four of these are at the happy-hour price. Can you make sure that stays on when it splits?”
“These add up to more than the check you showed us. I think the drink discount dropped off when it split.”
“The 20% was on the whole table — has that ended up only on one of these checks?”
“Since it’s a two-for-one, let’s keep it on one check and sort it out between us.”
The structural fix: divide the receipt, not the check
The failures above are not one failure, and they do not all come from the same place — Toast removes one kind of discount, Square relocates another and moves a third along with its item. What they share is narrower: each one happens because items were moved between checks. That is the operation with a cost, and it is worth separating from the operation people usually mean when they say “split the bill.”
Splitting the payment is not splitting the check. Both systems can take several payments against one intact check — Toast documents tendering a partial amount and leaving a balance due, and a Square community explanation of the BOGO failure notes that an even split “does not actually move items around. It just prorates the total.” Nothing moves, so nothing is dropped or stranded by the move itself. If the group is happy to divide by headcount, asking to split the payment sidesteps the item-move failures above — though neither vendor publishes a guarantee covering every promotion type, so a conditional deal is still worth asking about.
The catch is that headcount is exactly what people are trying to escape when one person had a $6 happy-hour beer and someone else had two $18 cocktails. The honest summary of the choice: an even payment split preserves the discount but not the fairness, and an item-level check split buys the fairness by asking the POS to rebuild the bill — which is where discounts get lost.
Splitting from the receipt is a way out of that trade. The restaurant’s check stays whole and gets paid once, so no discount is ever asked to survive being moved; the item-level division happens afterwards, on the document the restaurant already printed. The cost is real and worth naming: one card pays, so the group has to settle up between themselves rather than each paying the restaurant directly — which is the trade our comparison of splitting at the swipe versus at the receipt works through in full. What this page adds to that comparison is one specific entry: a check split at the terminal can cost you a discount you had already earned.
It is worth reading alongside how to tip when a discount is on the bill — that is the other half of the discount problem, and it runs in the opposite direction, because a discount that does survive changes what your tip should be calculated on.
FAQ
Common questions
01 Does splitting a restaurant check remove the discount?
It can, and it depends on the system and the type of discount. Toast's support documentation states that item-level discounts do not carry over when a check is split — the discount is removed from the item once it moves to the new check, and staff must reapply it on each split check individually. Square documents that item-specific discounts move alongside their item when splitting by seat, but that cart-level discounts and promotions remain on the original check unless moved manually.
02 Why did my happy-hour price disappear after the server split the bill?
Most likely because the discount was attached to the line item and the item was moved to a new check, which on some point-of-sale systems strips the discount. It is a documented behaviour rather than a mistake by the restaurant, but it does require the server to reapply the discount by hand. Ask for it to be corrected before the check is closed.
03 What happens to a BOGO or 'spend $50, get $10 off' deal when a check is split?
Conditional discounts test a relationship between items on one check. When the qualifying items are moved onto separate checks, the rule may no longer see a qualifying combination and the discount stops applying. This has been reported for one Square BOGO configuration when items were split across seats; how other systems and offer types behave is not documented. For conditional deals, the safest move is to keep the check whole and settle up between yourselves afterwards.
04 Who ends up with a whole-table discount when the check is split?
Square's documentation says that when a check is split by seat, cart-level discounts and promotions remain on the original check and can be moved manually. So the benefit stops being spread across the group: the new checks are created without it, and whoever holds the original keeps whatever it is still worth against the items left on that check. Staff can move it by hand, but generally only if someone asks.
05 How can I tell if a discount fell off when the check was split?
Add the split checks together and compare the sum against the combined check you were shown before the split. Each individual check will look internally correct, so the discrepancy only appears in the total. Photographing the combined check before the split gives you the reference number to compare against.
06 Is it better to split the check or split the payment?
It depends on what you are protecting. For loyalty points, Toast notes that guests earning points across multiple loyalty accounts on one order should have the check split rather than the payment, or points will not accrue properly. For item-level discounts on that same system, splitting the check is the operation that can strip them. If a discount is on the bill and the deal is conditional, the least lossy option is usually to leave the check intact and divide the money afterwards.
